Diffusion-prepared imaging with amplitude navigation for correction of motion-induced signal loss

Abstract

Purpose

Diffusion-prepared imaging is a flexible alternative to conventional spin-echo diffusion-weighted EPI that allows selection of different imaging readouts and k-space traversals, and permits control of image contrast or image artifacts. We investigate a new signal model and reconstruction for diffusion-prepared imaging that addresses signal variations caused by motion-sensitizing diffusion gradients.

Methods

A signal model, sampling theory, and reconstruction framework were developed assuming that motion-induced phases and the measured signals are random variables. The reconstruction incorporates real-valued amplitude weights estimated from low-resolution images into a linear system. A diffusion-prepared sequence was applied in phantom and in vivo acquisitions using different methods for managing phase errors from eddy currents or motion. This acquisition was performed with a high number of NEX and retrospectively undersampled to analyze errors in ADC estimation, and compared to spin-echo diffusion-weighted EPI, as well as conventional diffusion-prepared methods. The B₁ sensitivity of the sequence was investigated using simulation and phantom experiments.

Results

Images reconstructed using the proposed method had similar image structures when compared to conventional spin-echo diffusion-weighted EPI, and demonstrated improved SNR efficiency compared to previous diffusion-prepared approaches. ADC errors followed a trend consistent with the derived signal model, sampling theory, and expected B₁ sensitivity. The sampling requirement was shown to depend on the magnitude of motion-induced phases, as well as phases from residual eddy currents.

Conclusion

Employing amplitude weights in the reconstruction of a diffusion-prepared sequence can improve SNR efficiency at the cost of a greater minimum sampling time and increased sensitivity to B₁ inhomogeneity.